[PATCH 0/2] remove CGI.pm support

[PATCH 0/2] remove CGI.pm support

[Eric Wong]

We probably never should've supported CGI.pm in the first place,
but oh well.  We'll probably need to support running as .cgi
forever...

Eric Wong (2):
      remove direct CGI.pm support
      www: make interface more OO

 INSTALL                           |  5 +--
 Makefile.PL                       |  2 +-
 examples/public-inbox.psgi        |  7 ++--
 lib/PublicInbox/ExtMsg.pm         | 11 ++----
 lib/PublicInbox/Feed.pm           |  9 +----
 lib/PublicInbox/GitHTTPBackend.pm |  6 ++--
 lib/PublicInbox/WWW.pm            | 48 ++++++++++++-------------
 public-inbox-httpd                |  7 ++--
 public-inbox.cgi                  | 76 ++++++++++++---------------------------
 9 files changed, 56 insertions(+), 115 deletions(-)

[PATCH 1/2] remove direct CGI.pm support

[Eric Wong]

Relying on Plack::Handler::CGI is much easier for long-term
maintenance and development.

Nowadays, we even include our own httpd implementation to
facilitate easier deployment with PSGI/Plack.
---
 INSTALL                           |  5 +--
 Makefile.PL                       |  2 +-
 lib/PublicInbox/ExtMsg.pm         | 11 +-----
 lib/PublicInbox/Feed.pm           |  9 +----
 lib/PublicInbox/GitHTTPBackend.pm |  6 +--
 lib/PublicInbox/WWW.pm            | 24 ++----------
 public-inbox.cgi                  | 77 +++++++++++++--------------------------
 7 files changed, 36 insertions(+), 98 deletions(-)

diff --git a/INSTALL b/INSTALL
index 538a95d..54c344c 100644
--- a/INSTALL
+++ b/INSTALL
@@ -39,7 +39,7 @@ Requirements (server MDA)
 
 Optional modules:
 
-  - CGI[1]                     perl-modules[2]
+  - Plack[1]                   libplack-perl
   - Mail::Thread (2.5+)[1]     libmail-thread-perl
   - URI::Escape[1]             liburi-perl
   - Search::Xapian[3]          libsearch-xapian-perl
@@ -49,9 +49,6 @@ Optional modules:
   - Danga::Socket[4]           libdanga-socket-perl
 
 [1] - Only required for serving/generating Atom and HTML pages.
-[2] - Keep in mind this will be split into a separate Debian package
-      when CGI.pm is dropped from the Perl standard library.
-      Plack/PSGI and mod_perl2 are both supported.
 [3] - Optional for HTML web interface and NNTP server
 [4] - Optional for NNTP server
 
diff --git a/Makefile.PL b/Makefile.PL
index 5eec5ef..621baf6 100644
--- a/Makefile.PL
+++ b/Makefile.PL
@@ -17,7 +17,6 @@ WriteMakefile(
 		# note: we use ssoma(1) and spamc(1), NOT the Perl modules
 		# We also depend on git through ssoma.
 		# Keep this sorted and synced to the INSTALL document
-		'CGI' => 0,
 		'Date::Parse' => 0,
 		'Email::Address' => 0,
 		'Email::Filter' => 0,
@@ -27,6 +26,7 @@ WriteMakefile(
 		'File::Path::Expand' => 0,
 		'IPC::Run' => 0,
 		'Mail::Thread' => '2.5', # 2.5+ needed for Email::Simple compat
+		'Plack' => 0,
 		'URI::Escape' => 0,
 		# We have more test dependencies, but do not force
 		# users to install them
diff --git a/lib/PublicInbox/ExtMsg.pm b/lib/PublicInbox/ExtMsg.pm
index d89a7e3..cf9b6d5 100644
--- a/lib/PublicInbox/ExtMsg.pm
+++ b/lib/PublicInbox/ExtMsg.pm
@@ -86,15 +86,7 @@ sub ext_msg {
 	eval { require PublicInbox::Msgmap };
 	my $have_mm = $@ ? 0 : 1;
 	my $cgi = $ctx->{cgi};
-	my $base_url;
-	my $scheme;
-	if (ref($cgi) eq 'CGI') {
-		$base_url = $cgi->url(-base) . '/';
-		$scheme = $cgi->protocol;
-	} else { # Plack::Request
-		$base_url = $cgi->base->as_string;
-		$scheme = $cgi->env->{'psgi.url_scheme'};
-	}
+	my $base_url = $cgi->base->as_string;
 	if ($have_mm) {
 		my $tmp_mid = $mid;
 		my $url;
@@ -145,6 +137,7 @@ again:
 	if (@EXT_URL && index($mid, '@') >= 0) {
 		$code = 300;
 		$s .= "\nPerhaps try an external site:\n\n";
+		my $scheme = $cgi->scheme;
 		foreach my $u (@EXT_URL) {
 			$u = "$scheme:$u" if $u =~ m!\A//!;
 			my $r = sprintf($u, $href);
diff --git a/lib/PublicInbox/Feed.pm b/lib/PublicInbox/Feed.pm
index a0aa62a..a5828a8 100644
--- a/lib/PublicInbox/Feed.pm
+++ b/lib/PublicInbox/Feed.pm
@@ -284,14 +284,7 @@ sub get_feedopts {
 
 	my $url_base;
 	if ($cgi) {
-		my $base;
-		if (ref($cgi) eq 'CGI') {
-			$base = $cgi->url(-base);
-		} else { # Plack::Request
-			$base = $cgi->base->as_string;
-			$base =~ s!/\z!!;
-		}
-		$url_base = "$base/$listname";
+		$url_base = $cgi->base->as_string . $listname;
 		if (my $mid = $ctx->{mid}) { # per-thread feed:
 			$rv{atomurl} = "$url_base/$mid/t.atom";
 		} else {
diff --git a/lib/PublicInbox/GitHTTPBackend.pm b/lib/PublicInbox/GitHTTPBackend.pm
index 71b7a8f..562c290 100644
--- a/lib/PublicInbox/GitHTTPBackend.pm
+++ b/lib/PublicInbox/GitHTTPBackend.pm
@@ -55,7 +55,7 @@ sub serve {
 	my $len = $size;
 	my @h;
 
-	my $env = $cgi->{env} || \%ENV;
+	my $env = $cgi->{env};
 	my $range = $env->{HTTP_RANGE};
 	if (defined $range && $range =~ /\bbytes=(\d*)-(\d*)\z/) {
 		($code, $len) = prepare_range($cgi, $in, \@h, $1, $2, $size);
@@ -117,9 +117,7 @@ sub prepare_range {
 			push @$h, "bytes $beg-$end/$size";
 
 			# FIXME: Plack::Middleware::Deflater bug?
-			if (my $env = $cgi->{env}) {
-				$env->{'psgix.no-compress'} = 1;
-			}
+			$cgi->{env}->{'psgix.no-compress'} = 1;
 		}
 	}
 	($code, $len);
diff --git a/lib/PublicInbox/WWW.pm b/lib/PublicInbox/WWW.pm
index 1f28df2..651c19e 100644
--- a/lib/PublicInbox/WWW.pm
+++ b/lib/PublicInbox/WWW.pm
@@ -223,12 +223,6 @@ sub get_thread {
 	PublicInbox::View::thread_html($ctx, $foot, $srch);
 }
 
-sub self_url {
-	my ($cgi) = @_;
-						# Plack::Request
-	ref($cgi) eq 'CGI' ? $cgi->self_url : $cgi->uri->as_string;
-}
-
 sub ctx_get {
 	my ($ctx, $key) = @_;
 	my $val = $ctx->{$key};
@@ -258,9 +252,7 @@ sub footer {
 	my $urls = try_cat("$git_dir/cloneurl");
 	my @urls = split(/\r?\n/, $urls || '');
 	my %seen = map { $_ => 1 } @urls;
-	my $cgi = $ctx->{cgi};
-	my $http = (ref($cgi) eq 'CGI') ? $cgi->url(-base) . "/$listname" :
-			$cgi->base->as_string . $listname;
+	my $http = $ctx->{cgi}->base->as_string . $listname;
 	$seen{$http} or unshift @urls, $http;
 	if (scalar(@urls) == 1) {
 		$urls = "URL for <a\nhref=\"" . SSOMA_URL .
@@ -325,7 +317,7 @@ sub get_thread_mbox {
 sub get_thread_atom {
 	my ($ctx) = @_;
 	searcher($ctx) or return need_search($ctx);
-	$ctx->{self_url} = self_url($ctx->{cgi});
+	$ctx->{self_url} = $ctx->{cgi}->uri->as_string;
 	require PublicInbox::Feed;
 	PublicInbox::Feed::generate_thread_atom($ctx);
 }
@@ -390,16 +382,8 @@ sub r301 {
 	my ($ctx, $listname, $mid, $suffix) = @_;
 	my $cgi = $ctx->{cgi};
 	my $url;
-	my $qs;
-	if (ref($cgi) eq 'CGI') {
-		$url = $cgi->url(-base) . '/';
-		$qs = $cgi->query_string;
-	} else { # Plack::Request
-		$url = $cgi->base->as_string;
-		$qs = $cgi->env->{QUERY_STRING};
-	}
-
-	$url .= $listname . '/';
+	my $qs = $cgi->env->{QUERY_STRING};
+	$url = $cgi->base->as_string . $listname . '/';
 	$url .= (uri_escape_utf8($mid) . '/') if (defined $mid);
 	$url .= $suffix if (defined $suffix);
 	$url .= "?$qs" if $qs ne '';
diff --git a/public-inbox.cgi b/public-inbox.cgi
index 4b74a62..e73e23c 100755
--- a/public-inbox.cgi
+++ b/public-inbox.cgi
@@ -1,62 +1,35 @@
 #!/usr/bin/perl -w
-# Copyright (C) 2014-2015 all contributors <meta@public-inbox.org>
-# License: AGPLv3 or later (https://www.gnu.org/licenses/agpl-3.0.txt)
+# Copyright (C) 2014-2016 all contributors <meta@public-inbox.org>
+# License: AGPL-3.0+ or later <https://www.gnu.org/licenses/agpl-3.0.txt>
 #
 # Enables using PublicInbox::WWW as a CGI script
 use strict;
 use warnings;
-use IO::Handle;
+use Plack::Loader;
+use Plack::Builder;
+use Plack::Request;
+use Plack::Handler::CGI;
 use PublicInbox::WWW;
-use CGI qw/-nosticky/;
-our $NO_SCRIPT_NAME;
-our %HTTP_CODES;
-BEGIN {
-	$NO_SCRIPT_NAME = 1 if $ENV{NO_SCRIPT_NAME};
-	if ($ENV{MOD_PERL}) {
-		CGI->compile;
-		PublicInbox::WWW->preload;
-	}
-
-	%HTTP_CODES = (
-		200 => 'OK',
-		300 => 'Multiple Choices',
-		301 => 'Moved Permanently',
-		302 => 'Found',
-		404 => 'Not Found',
-		405 => 'Method Not Allowed',
-		501 => 'Not Implemented',
-	);
-}
-
-# some servers (Ruby webrick) include scheme://host[:port] here,
-# which confuses CGI.pm when generating self_url.
-# RFC 3875 does not mention REQUEST_URI at all,
-# so nuke it since CGI.pm functions without it.
-delete $ENV{REQUEST_URI};
-$ENV{SCRIPT_NAME} = '' if $NO_SCRIPT_NAME;
-my $req = CGI->new;
-my $ret = PublicInbox::WWW::run($req, $req->request_method);
+BEGIN { PublicInbox::WWW->preload if $ENV{MOD_PERL} }
 
-my $out = select;
-$out->binmode;
-
-if (ref($ret) eq 'CODE') {
-	$ret->(*dump_header);
-} else {
-	my ($status, $headers, $body) = @$ret;
+my $have_deflater = eval { require Plack::Middleware::Deflater; 1 };
+my $app = builder {
+	if ($have_deflater) {
+		enable 'Deflater',
+			content_type => [ 'text/html', 'text/plain',
+					'application/atom+xml' ];
+	}
 
-	dump_header([$status, $headers])->write($body->[0]);
-}
+	# Enable to ensure redirects and Atom feed URLs are generated
+	# properly when running behind a reverse proxy server which
+	# sets X-Forwarded-For and X-Forwarded-Proto request headers.
+	# See Plack::Middleware::ReverseProxy documentation for details
+	# enable 'ReverseProxy';
 
-sub dump_header {
-	my ($res) = @_;
-	my $fh = select;
-	my ($status, $headers) = @$res;
-	$fh->write("Status: $status $HTTP_CODES{$status}\r\n");
-	my @tmp = @$headers;
-	while (my ($k, $v) = splice(@tmp, 0, 2)) {
-		$fh->write("$k: $v\r\n");
+	enable 'Head';
+	sub {
+		my $req = Plack::Request->new(@_);
+		PublicInbox::WWW::run($req, $req->method);
 	}
-	$fh->write("\r\n");
-	$fh;
-}
+};
+Plack::Handler::CGI->new->run($app);
-- 
EW

[PATCH 2/2] www: make interface more OO

[Eric Wong]

This allows multiple instances the WWW app from
running within the same process space
---
 examples/public-inbox.psgi |  7 ++-----
 lib/PublicInbox/WWW.pm     | 24 ++++++++++++++++++------
 public-inbox-httpd         |  7 ++-----
 public-inbox.cgi           |  7 ++-----
 4 files changed, 24 insertions(+), 21 deletions(-)

diff --git a/examples/public-inbox.psgi b/examples/public-inbox.psgi
index c4a8903..e3e4215 100644
--- a/examples/public-inbox.psgi
+++ b/examples/public-inbox.psgi
@@ -10,7 +10,7 @@ PublicInbox::WWW->preload;
 use Plack::Request;
 use Plack::Builder;
 my $have_deflater = eval { require Plack::Middleware::Deflater; 1 };
-
+my $www = PublicInbox::WWW->new;
 builder {
 	enable 'Chunked';
 	if ($have_deflater) {
@@ -25,8 +25,5 @@ builder {
 	# See Plack::Middleware::ReverseProxy documentation for details
 	# enable 'ReverseProxy';
 	enable 'Head';
-	sub {
-		my $req = Plack::Request->new(@_);
-		PublicInbox::WWW::run($req, $req->method);
-	}
+	sub { $www->call(@_) };
 }
diff --git a/lib/PublicInbox/WWW.pm b/lib/PublicInbox/WWW.pm
index 651c19e..e87e559 100644
--- a/lib/PublicInbox/WWW.pm
+++ b/lib/PublicInbox/WWW.pm
@@ -22,14 +22,26 @@ use PublicInbox::GitHTTPBackend;
 our $LISTNAME_RE = qr!\A/([\w\.\-]+)!;
 our $MID_RE = qr!([^/]+)!;
 our $END_RE = qr!(f/|T/|t/|R/|t\.mbox(?:\.gz)?|t\.atom|raw|)!;
-our $pi_config;
 
-sub run {
-	my ($cgi, $method) = @_;
+sub new {
+	my ($class, $pi_config) = @_;
 	$pi_config ||= PublicInbox::Config->new;
-	my $ctx = { cgi => $cgi, pi_config => $pi_config };
+	bless { pi_config => $pi_config }, $class;
+}
+
+# backwards compatibility, do not use
+sub run {
+	my ($req, $method) = @_;
+	PublicInbox::WWW->new->call($req->env);
+}
+
+sub call {
+	my ($self, $env) = @_;
+	my $cgi = Plack::Request->new($env);
+	my $ctx = { cgi => $cgi, pi_config => $self->{pi_config} };
 	my $path_info = $cgi->path_info;
 
+	my $method = $cgi->method;
 	if ($method eq 'POST' &&
 		 $path_info =~ m!$LISTNAME_RE/(git-upload-pack)\z!) {
 		my $path = $2;
@@ -107,7 +119,7 @@ sub r { [ $_[0], ['Content-Type' => 'text/plain'], [ join(' ', @_, "\n") ] ] }
 # returns undef if valid, array ref response if invalid
 sub invalid_list {
 	my ($ctx, $listname) = @_;
-	my $git_dir = $pi_config->get($listname, "mainrepo");
+	my $git_dir = $ctx->{pi_config}->get($listname, "mainrepo");
 	if (defined $git_dir) {
 		$ctx->{git_dir} = $git_dir;
 		$ctx->{git} = PublicInbox::Git->new($git_dir);
@@ -264,7 +276,7 @@ sub footer {
 			join("\n", map { "\t$_" } @urls);
 	}
 
-	my $addr = $pi_config->get($listname, 'address');
+	my $addr = $ctx->{pi_config}->get($listname, 'address');
 	if (ref($addr) eq 'ARRAY') {
 		$addr = $addr->[0]; # first address is primary
 	}
diff --git a/public-inbox-httpd b/public-inbox-httpd
index 6436bd7..3635c9a 100644
--- a/public-inbox-httpd
+++ b/public-inbox-httpd
@@ -24,6 +24,7 @@ my $refresh = sub {
 "$0 runs in /, command-line paths must be absolute\n";
 		}
 	} else {
+		my $www = PublicInbox::WWW->new;
 		$app = eval {
 			my $deflate_types = eval {
 				require Plack::Middleware::Deflater;
@@ -37,11 +38,7 @@ my $refresh = sub {
 						content_type => $deflate_types
 				}
 				enable 'Head';
-				sub {
-					my $req = Plack::Request->new(@_);
-					PublicInbox::WWW::run($req,
-							$req->method);
-				};
+				sub { $www->call(@_) };
 			};
 		};
 	}
diff --git a/public-inbox.cgi b/public-inbox.cgi
index e73e23c..ee9510c 100755
--- a/public-inbox.cgi
+++ b/public-inbox.cgi
@@ -11,7 +11,7 @@ use Plack::Request;
 use Plack::Handler::CGI;
 use PublicInbox::WWW;
 BEGIN { PublicInbox::WWW->preload if $ENV{MOD_PERL} }
-
+my $www = PublicInbox::WWW->new;
 my $have_deflater = eval { require Plack::Middleware::Deflater; 1 };
 my $app = builder {
 	if ($have_deflater) {
@@ -27,9 +27,6 @@ my $app = builder {
 	# enable 'ReverseProxy';
 
 	enable 'Head';
-	sub {
-		my $req = Plack::Request->new(@_);
-		PublicInbox::WWW::run($req, $req->method);
-	}
+	sub { $www->call(@_) };
 };
 Plack::Handler::CGI->new->run($app);
-- 
EW